Ball Detection Based on Color Information and Hough Transform

Abstract

In this article, we present a ball detection algorithm based on color information and Hough transform. Our objective is to detect the color balls under complex environments and different illumination conditions. To do so, we firstly apply the multiscale Retinex algorithm to enhance the image. Then, an adaptive Hough transform is introduced to locate the position of the balls in order to reduce the influence of the complex environments. Thirdly, to detect the interested balls under different illumination conditions, we introduce color invariant model Grayworld normalization (GN) to overcome the influence of the illumination. Finally, to evaluate the proposed ball detection in practice, experiments have been carried out on 100 images taken from various illuminations. On the basis of the reported theory and experimental results, it is shown that high interested ball detection accuracy is achieved by our proposed method.